Computation on architectures that feature fine-grained parallelism requires algorithms that overcome load imbalance, inefficient memory access, serialization, and excessive synchronization. In this ...